Recently, 3D city modelling has attracted a growing interest as a building block for creating city Digital Twins. They are complex representations that include interactive representations of buildings and infrastructures, integrated with the wide range of data typically useful in a Smart City environment. This paper presents an automatic method for producing 3D city models from a various set of data, as well as its integration into the open-source Smart City framework, Snap4City. The proposed solution offers a method for creating effective integrated data visualizations of 3D city entities coupled with a large variety of Smart City data (eg, IoT Devices which generate time-series data, heatmaps, geometries and shapes related to traffic flows, bus routes, cycling paths). The solution is based on a deep learning approach for rooftop detection and alignment based on a U-Net architecture. The implementation has been enforced into the open-source Snap4City Smart City platform, and has been validated by using a manually created ground-truth of 200 buildings scattered uniformly in the central area of Florence, plus a number of meshes representing a number of facades (not detailed in this paper), and traffic flows, pins, heatmaps, etc.